The A.I. Turducken

AI Turducken.jpg

A.I. literacy is quickly rising in the technical pecking order to become a must-have feather in the cap of developers, professionals, and entrepreneurs of all stripes.  Like any new field enduring a sudden rise in notoriety, the terminology of A.I. has often run afoul of misuse, misunderstanding, and general henpecking at the hands and pens of hasty bloggers. In this article we’ll provide a simple explanation of three essential terms in the field: Artificial Intelligence, Machine Leaning, and Deep Learning.

Artificial Intelligence (A.I.) is our big turkey. Essentially, A.I. is any type of program that asks a computer to think or act more like a person (or animal). A.I. has been theorized about for a long time, but has often remained just beyond the grasp of researchers. Computers are, after all, very different from living things. While most turkeys are too birdbrained to do basic math, they can almost immediately tell the difference between a tasty bug and an inedible rock. Computers on the other hand are excellent at math, but until recently were not very good at bug vs. rock identification. Thanks to our second bird, that’s begun to change.

Machine Learning is our fat duck. Nestled inside A.I., machine learning is a process of getting a computer to learn how to do something rather than telling it exactly what to do. This is accomplished through the clever use of statistics. Think about the spam filter on your email account, how does it know the difference between a real email and junk? The simple answer is, you’ve helped train it to do so.

Every time you’ve marked something as spam you’ve been helping improve an algorithm that does that work for you. If a statistically relevant number of emails with the title “The Prince of Norway Wants to Give You Money” from royalty@giantscam.com are marked as spam, the spam filter will begin to catch wise. Eventually it will start to block emails with a title like “The [Title] of [Place Name] Wants to Give You [Something]” from [Anybody]@giantscam.com. The spam filter is not blocking those emails because it was explicitly told to do so, it’s doing so because people helped it learn how through their behavior. More correctly labeled emails (data) leads to better, smarter algorithms. That’s machine learning and it’s just ducky, but very recently learning algorithms have gone even deeper. 

Deep Learning, the clever chicken nested at the center of our feathered trio, is the secret ingredient for making really smart A.I.. A subset of machine learning, deep learning is poised to fundamentally change the way people work, live, and even think. It’s an enormously powerful discovery in computer science, but it’s not terribly difficult to understand or even to code.

Have you ever pulled your hand away from something hot without thinking about it? That simple, instinctive reaction is hardwired in your mind and body. You don’t need to consider the pros and cons of leaving your hand on a flame, you simply pull away because that’s what your nervous system is biologically programed to do.

With enough data and processing power, deep learning frameworks like Tensorflow, Caffe, and PyTorch allow data scientists to create what’s called neural networks which take in tons of information and output desired results. How they do this is no less complex than the series of steps your nervous system takes to pull your hand away from a fire. We can only really understand neural networks after they’ve accomplished what we’ve asked of them using a process called backpropagation.

Deep learning technology is helping us create programs that can drive cars, identify health problems, and even search for habitable planets in space. These programs can perceive of things that no human can. That’s one smart chicken. What’s truly incredible about deep learning frameworks is that they can create customized neural networks on your home computer in a relatively short period of time (a decent graphics card helps).

To recap:

·       The Turkey: A.I. is a program that asks a computer to be more like a brain

·       The Duck: Machine Learning, is the ability of an A.I. to learn how to do something rather than be told

·       The Chicken: Deep Learning is a type of machine learning that uses big data to create neural networks, essentially digital instincts for an A.I.

To learn more about A.I. we suggest you check out Siraj Raval’s YouTube channel, The Deep Learning Book, and Udacity’s free course: Intro to Machine Learning.

If you have need for professional consulting in machine learning, feel free to contact us directly. If we can’t help answer your questions, we can at least point you in the right direction.

Thanks for checking out our blog. Stay Hungry!